This paper examines the short-run and long-run relationships between visitor arrivals in Fiji, real disposable incomes and relative hotel and substitute prices for the period 1970-2000, using cointegration techniques and error correction models. The paper uses a cointegration technique - the bounds testing approach developed within an autoregressive distributed lag (ARDL) framework - that has not previously been used to estimate tourism demand models. The main advantage of the approach is that, apart from providing robust results in small sample sizes, it needs no a priori knowledge about the integration properties of the variables. The long-run results indicate that growth in income in Fiji's main tourist source markets has a positive impact on visitor arrivals while relative hotel and substitute prices have negative impacts on visitor arrivals. The findings are consistent with economic theory and proffer important policy implications.